microRNAs (miRs, miRNAs) are a novel class of non-coding, regulatory RNA genes1-3 which are involved in oncogenesis4 and show remarkable tissue-specificity5-7. They have emerged as highly tissue-specific biomarkers2′5′6 postulated to play important roles in encoding developmental decisions of differentiation. Various studies have tied microRNAs to the development of specific malignancies4. MicroRNAs are also stable in tissue, stored frozen or as formalin-fixed, paraffin-embedded (FFPE) samples, and in serum.
Hundreds of thousands of patients in the U.S. are diagnosed each year with a cancer that has already metastasized, without a clearly identified primary site. Oncologists and pathologists are constantly faced with a diagnostic dilemma when trying to identify the primary origin of a patient's metastasis. As metastases need to be treated according to their primary origin, accurate identification of the metastases' primary origin can be critical for determining appropriate treatment.
Once a metastatic tumor is found, the patient may undergo a wide range of costly, time consuming, and at times inefficient tests, including physical examination of the patient, histopathology analysis of the biopsy, imaging methods such as chest X-ray, CT and PET scans, in order to identify the primary origin of the metastasis.
Metastatic cancer of unknown primary (CUP) accounts for 3-5% of all new cancer cases, and as a group is usually a very aggressive disease with a poor prognosis10. The concept of CUP comes from the limitation of present methods to identify cancer origin, despite an often complicated and costly process which can significantly delay proper treatment of such patients. Recent studies revealed a high degree of variation in clinical management, in the absence of evidence based treatment for CUP11. Many protocols were evaluated12 but have shown relatively small benefit13. Determining tumor tissue of origin is thus an important clinical application of molecular diagnostics9.
Molecular classification studies for tumor tissue origin14-17 have generally used classification algorithms that did not utilize domain-specific knowledge: tissues were treated as a-priori equivalents, ignoring underlying similarities between tissue types with a common developmental origin in embryogenesis. An exception of note is the study by Shedden and co-workers18, that was based on a pathology classification tree. These studies used machine-learning methods that average effects of biological features (e.g., mRNA expression levels), an approach which is more amenable to automated processing but does not use or generate mechanistic insights.
Various markers have been proposed to indicate specific types of cancers and tumor tissue of origin. However, the diagnostic accuracy of tumor markers has not yet been defined. There is thus a need for a more efficient and effective method for diagnosing and classifying specific types of cancers.